spandsp  0.0.6
v17rx.h
Go to the documentation of this file.
1 /*
2  * SpanDSP - a series of DSP components for telephony
3  *
4  * v17rx.h - ITU V.17 modem receive part
5  *
6  * Written by Steve Underwood <steveu@coppice.org>
7  *
8  * Copyright (C) 2003 Steve Underwood
9  *
10  * All rights reserved.
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU Lesser General Public License version 2.1,
14  * as published by the Free Software Foundation.
15  *
16  * This program is distributed in the hope that it will be useful,
17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19  * GNU Lesser General Public License for more details.
20  *
21  * You should have received a copy of the GNU Lesser General Public
22  * License along with this program; if not, write to the Free Software
23  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
24  */
25 
26 /*! \file */
27 
28 #if !defined(_SPANDSP_V17RX_H_)
29 #define _SPANDSP_V17RX_H_
30 
31 /*! \page v17rx_page The V.17 receiver
32 \section v17rx_page_sec_1 What does it do?
33 The V.17 receiver implements the receive side of a V.17 modem. This can operate
34 at data rates of 14400, 12000, 9600 and 7200 bits/second. The audio input is a stream
35 of 16 bit samples, at 8000 samples/second. The transmit and receive side of V.17
36 modems operate independantly. V.17 is mostly used for FAX transmission over PSTN
37 lines, where it provides the standard 14400 bits/second rate.
38 
39 \section v17rx_page_sec_2 How does it work?
40 V.17 uses QAM modulation, at 2400 baud, and trellis coding. Constellations with
41 16, 32, 64, and 128 points are defined. After one bit per baud is absorbed by the
42 trellis coding, this gives usable bit rates of 7200, 9600, 12000, and 14400 per
43 second.
44 
45 V.17 specifies a training sequence at the start of transmission, which makes the
46 design of a V.17 receiver relatively straightforward. The first stage of the
47 training sequence consists of 256
48 symbols, alternating between two constellation positions. The receiver monitors
49 the signal power, to sense the possible presence of a valid carrier. When the
50 alternating signal begins, the power rising above a minimum threshold (-43dBm0)
51 causes the main receiver computation to begin. The initial measured power is
52 used to quickly set the gain of the receiver. After this initial settling, the
53 front end gain is locked, and the adaptive equalizer tracks any subsequent
54 signal level variation. The signal is oversampled to 24000 samples/second (i.e.
55 signal, zero, zero, signal, zero, zero, ...) and fed to a complex root raised
56 cosine pulse shaping filter. This filter has been modified from the conventional
57 root raised cosine filter, by shifting it up the band, to be centred at the nominal
58 carrier frequency. This filter interpolates the samples, pulse shapes, and performs
59 a fractional sample delay at the same time. 192 sets of filter coefficients are used
60 to achieve a set of finely spaces fractional sample delays, between zero and
61 one sample. By choosing every fifth sample, and the appropriate set of filter
62 coefficients, the properly tuned symbol tracker can select data samples at 4800
63 samples/second from points within 0.28 degrees of the centre and mid-points of
64 each symbol. The output of the filter is multiplied by a complex carrier, generated
65 by a DDS. The result is a baseband signal, requiring no further filtering, apart from
66 an adaptive equalizer. The baseband signal is fed to a T/2 adaptive equalizer.
67 A band edge component maximisation algorithm is used to tune the sampling, so the samples
68 fed to the equalizer are close to the mid point and edges of each symbol. Initially
69 the algorithm is very lightly damped, to ensure the symbol alignment pulls in
70 quickly. Because the sampling rate will not be precisely the same as the
71 transmitter's (the spec. says the symbol timing should be within 0.01%), the
72 receiver constantly evaluates and corrects this sampling throughout its
73 operation. During the symbol timing maintainence phase, the algorithm uses
74 a heavier damping.
75 
76 The carrier is specified as 1800Hz +- 1Hz at the transmitter, and 1800 +-7Hz at
77 the receiver. The receive carrier would only be this inaccurate if the link
78 includes FDM sections. These are being phased out, but the design must still
79 allow for the worst case. Using an initial 1800Hz signal for demodulation gives
80 a worst case rotation rate for the constellation of about one degree per symbol.
81 Once the symbol timing synchronisation algorithm has been given time to lock to the
82 symbol timing of the initial alternating pattern, the phase of the demodulated signal
83 is recorded on two successive symbols - once for each of the constellation positions.
84 The receiver then tracks the symbol alternations, until a large phase jump occurs.
85 This signifies the start of the next phase of the training sequence. At this
86 point the total phase shift between the original recorded symbol phase, and the
87 symbol phase just before the phase jump occurred is used to provide a coarse
88 estimation of the rotation rate of the constellation, and it current absolute
89 angle of rotation. These are used to update the current carrier phase and phase
90 update rate in the carrier DDS. The working data already in the pulse shaping
91 filter and equalizer buffers is given a similar step rotation to pull it all
92 into line. From this point on, a heavily damped integrate and dump approach,
93 based on the angular difference between each received constellation position and
94 its expected position, is sufficient to track the carrier, and maintain phase
95 alignment. A fast rough approximator for the arc-tangent function is adequate
96 for the estimation of the angular error.
97 
98 The next phase of the training sequence is a scrambled sequence of two
99 particular symbols. We train the T/2 adaptive equalizer using this sequence. The
100 scrambling makes the signal sufficiently diverse to ensure the equalizer
101 converges to the proper generalised solution. At the end of this sequence, the
102 equalizer should be sufficiently well adapted that is can correctly resolve the
103 full QAM constellation. However, the equalizer continues to adapt throughout
104 operation of the modem, fine tuning on the more complex data patterns of the
105 full QAM constellation.
106 
107 In the last phase of the training sequence, the modem enters normal data
108 operation, with a short defined period of all ones as data. As in most high
109 speed modems, data in a V.17 modem passes through a scrambler, to whiten the
110 spectrum of the signal. The transmitter should initialise its data scrambler,
111 and pass the ones through it. At the end of the ones, real data begins to pass
112 through the scrambler, and the transmit modem is in normal operation. The
113 receiver tests that ones are really received, in order to verify the modem
114 trained correctly. If all is well, the data following the ones is fed to the
115 application, and the receive modem is up and running. Unfortunately, some
116 transmit side of some real V.17 modems fail to initialise their scrambler before
117 sending the ones. This means the first 23 received bits (the length of the
118 scrambler register) cannot be trusted for the test. The receive modem,
119 therefore, only tests that bits starting at bit 24 are really ones.
120 
121 The V.17 signal is trellis coded. Two bits of each symbol are convolutionally coded
122 to form a 3 bit trellis code - the two original bits, plus an extra redundant bit. It
123 is possible to ignore the trellis coding, and just decode the non-redundant bits.
124 However, the noise performance of the receiver would suffer. Using a proper
125 trellis decoder adds several dB to the noise tolerance to the receiving modem. Trellis
126 coding seems quite complex at first sight, but is fairly straightforward once you
127 get to grips with it.
128 
129 Trellis decoding tracks the data in terms of the possible states of the convolutional
130 coder at the transmitter. There are 8 possible states of the V.17 coder. The first
131 step in trellis decoding is to find the best candidate constellation point
132 for each of these 8 states. One of thse will be our final answer. The constellation
133 has been designed so groups of 8 are spread fairly evenly across it. Locating them
134 is achieved is a reasonably fast manner, by looking up the answers in a set of space
135 map tables. The disadvantage is the tables are potentially large enough to affect
136 cache performance. The trellis decoder works over 16 successive symbols. The result
137 of decoding is not known until 16 symbols after the data enters the decoder. The
138 minimum total accumulated mismatch between each received point and the actual
139 constellation (termed the distance) is assessed for each of the 8 states. A little
140 analysis of the coder shows that each of the 8 current states could be arrived at
141 from 4 different previous states, through 4 different constellation bit patterns.
142 For each new state, the running total distance is arrived at by inspecting a previous
143 total plus a new distance for the appropriate 4 previous states. The minimum of the 4
144 values becomes the new distance for the state. Clearly, a mechanism is needed to stop
145 this distance from growing indefinitely. A sliding window, and several other schemes
146 are possible. However, a simple single pole IIR is very simple, and provides adequate
147 results.
148 
149 For each new state we store the constellation bit pattern, or path, to that state, and
150 the number of the previous state. We find the minimum distance amongst the 8 new
151 states for each new symbol. We then trace back through the states, until we reach the
152 one 16 states ago which leads to the current minimum distance. The bit pattern stored
153 there is the error corrected bit pattern for that symbol.
154 
155 So, what does Trellis coding actually achieve? TCM is easier to understand by looking
156 at the V.23bis modem spec. The V.32bis spec. is very similar to V.17, except that it
157 is a full duplex modem and has non-TCM options, as well as the TCM ones in V.17.
158 
159 V32bis defines two options for pumping 9600 bits per second down a phone line - one
160 with and one without TCM. Both run at 2400 baud. The non-TCM one uses simple 16 point
161 QAM on the raw data. The other takes two out of every four raw bits, and convolutionally
162 encodes them to 3. Now we have 5 bits per symbol, and we need 32 point QAM to send the
163 data.
164 
165 The raw error rate from simple decoding of the 32 point QAM is horrible compared to
166 decoding the 16 point QAM. If a point decoded from the 32 point QAM is wrong, the likely
167 correct choice should be one of the adjacent ones. It is unlikely to have been one that
168 is far away across the constellation, unless there was a huge noise spike, interference,
169 or something equally nasty. Now, the 32 point symbols do not exist in isolation. There
170 was a kind of temporal smearing in the convolutional coding. It created a well defined
171 dependency between successive symbols. If we knew for sure what the last few symbols
172 were, they would lead us to a limited group of possible values for the current symbol,
173 constrained by the behaviour of the convolutional coder. If you look at how the symbols
174 were mapped to constellation points, you will see the mapping tries to spread those
175 possible symbols as far apart as possible. This will leave only one that is pretty
176 close to the received point, which must be the correct choice. However, this assumes
177 we know the last few symbols for sure. Since we don't, we have a bit more work to do
178 to achieve reliable decoding.
179 
180 Instead of decoding to the nearest point on the constellation, we decode to a group of
181 likely constellation points in the neighbourhood of the received point. We record the
182 mismatch for each - that is the distance across the constellation between the received
183 point and the group of nearby points. To avoid square roots, recording x2 + y2 can be
184 good enough. Symbol by symbol, we record this information. After a few symbols we can
185 stand back and look at the recorded information.
186 
187 For each symbol we have a set of possible symbol values and error metric pairs. The
188 dependency between symbols, created by the convolutional coder, means some paths from
189 symbol to symbol are possible and some are not. It we trace back through the possible
190 symbol to symbol paths, and total up the error metric through those paths, we end up
191 with a set of figures of merit (or more accurately figures of demerit, since
192 larger == worse) for the likelihood of each path being the correct one. The path with
193 the lowest total metric is the most likely, and gives us our final choice for what we
194 think the current symbol really is.
195 
196 That was hard work. It takes considerable computation to do this selection and traceback,
197 symbol by symbol. We need to get quite a lot from this. It needs to drive the error rate
198 down so far that is compensates for the much higher error rate due to the larger
199 constellation, and then buys us some actual benefit. Well in the example we are looking
200 at - V.32bis at 9600bps - it works out the error rate from the TCM option is like using
201 the non-TCM option with several dB more signal to noise ratio. That's nice. The non-TCM
202 option is pretty reasonable on most phone lines, but a better error rate is always a
203 good thing. However, V32bis includes a 14,400bps option. That uses 2400 baud, and 6 bit
204 symbols. Convolutional encoding increases that to 7 bits per symbol, by taking 2 bits and
205 encoding them to 3. This give a 128 point QAM constellation. Again, the difference between
206 using this, and using just an uncoded 64 point constellation is equivalent to maybe 5dB of
207 extra signal to noise ratio. However, in this case it is the difference between the modem
208 working only on the most optimal lines, and being widely usable across most phone lines.
209 TCM absolutely transformed the phone line modem business.
210 */
211 
212 /*!
213  V.17 modem receive side descriptor. This defines the working state for a
214  single instance of a V.17 modem receiver.
215 */
217 
218 #if defined(__cplusplus)
219 extern "C"
220 {
221 #endif
222 
223 /*! Initialise a V.17 modem receive context.
224  \brief Initialise a V.17 modem receive context.
225  \param s The modem context.
226  \param bit_rate The bit rate of the modem. Valid values are 7200, 9600, 12000 and 14400.
227  \param put_bit The callback routine used to put the received data.
228  \param user_data An opaque pointer passed to the put_bit routine.
229  \return A pointer to the modem context, or NULL if there was a problem. */
230 SPAN_DECLARE(v17_rx_state_t *) v17_rx_init(v17_rx_state_t *s, int bit_rate, put_bit_func_t put_bit, void *user_data);
231 
232 /*! Reinitialise an existing V.17 modem receive context.
233  \brief Reinitialise an existing V.17 modem receive context.
234  \param s The modem context.
235  \param bit_rate The bit rate of the modem. Valid values are 7200, 9600, 12000 and 14400.
236  \param short_train TRUE if a short training sequence is expected.
237  \return 0 for OK, -1 for bad parameter */
238 SPAN_DECLARE(int) v17_rx_restart(v17_rx_state_t *s, int bit_rate, int short_train);
239 
240 /*! Release a V.17 modem receive context.
241  \brief Release a V.17 modem receive context.
242  \param s The modem context.
243  \return 0 for OK */
244 SPAN_DECLARE(int) v17_rx_release(v17_rx_state_t *s);
245 
246 /*! Free a V.17 modem receive context.
247  \brief Free a V.17 modem receive context.
248  \param s The modem context.
249  \return 0 for OK */
250 SPAN_DECLARE(int) v17_rx_free(v17_rx_state_t *s);
251 
252 /*! Get the logging context associated with a V.17 modem receive context.
253  \brief Get the logging context associated with a V.17 modem receive context.
254  \param s The modem context.
255  \return A pointer to the logging context */
257 
258 /*! Change the put_bit function associated with a V.17 modem receive context.
259  \brief Change the put_bit function associated with a V.17 modem receive context.
260  \param s The modem context.
261  \param put_bit The callback routine used to handle received bits.
262  \param user_data An opaque pointer. */
263 SPAN_DECLARE(void) v17_rx_set_put_bit(v17_rx_state_t *s, put_bit_func_t put_bit, void *user_data);
264 
265 /*! Change the modem status report function associated with a V.17 modem receive context.
266  \brief Change the modem status report function associated with a V.17 modem receive context.
267  \param s The modem context.
268  \param handler The callback routine used to report modem status changes.
269  \param user_data An opaque pointer. */
270 SPAN_DECLARE(void) v17_rx_set_modem_status_handler(v17_rx_state_t *s, modem_status_func_t handler, void *user_data);
271 
272 /*! Process a block of received V.17 modem audio samples.
273  \brief Process a block of received V.17 modem audio samples.
274  \param s The modem context.
275  \param amp The audio sample buffer.
276  \param len The number of samples in the buffer.
277  \return The number of samples unprocessed.
278 */
279 SPAN_DECLARE_NONSTD(int) v17_rx(v17_rx_state_t *s, const int16_t amp[], int len);
280 
281 /*! Fake processing of a missing block of received V.17 modem audio samples.
282  (e.g due to packet loss).
283  \brief Fake processing of a missing block of received V.17 modem audio samples.
284  \param s The modem context.
285  \param len The number of samples to fake.
286  \return The number of samples unprocessed.
287 */
288 SPAN_DECLARE_NONSTD(int) v17_rx_fillin(v17_rx_state_t *s, int len);
289 
290 /*! Get a snapshot of the current equalizer coefficients.
291  \brief Get a snapshot of the current equalizer coefficients.
292  \param s The modem context.
293  \param coeffs The vector of complex coefficients.
294  \return The number of coefficients in the vector. */
295 #if defined(SPANDSP_USE_FIXED_POINTx)
296 SPAN_DECLARE(int) v17_rx_equalizer_state(v17_rx_state_t *s, complexi_t **coeffs);
297 #else
298 SPAN_DECLARE(int) v17_rx_equalizer_state(v17_rx_state_t *s, complexf_t **coeffs);
299 #endif
300 
301 /*! Get the current received carrier frequency.
302  \param s The modem context.
303  \return The frequency, in Hertz. */
304 SPAN_DECLARE(float) v17_rx_carrier_frequency(v17_rx_state_t *s);
305 
306 /*! Get the current symbol timing correction since startup.
307  \param s The modem context.
308  \return The correction. */
309 SPAN_DECLARE(float) v17_rx_symbol_timing_correction(v17_rx_state_t *s);
310 
311 /*! Get a current received signal power.
312  \param s The modem context.
313  \return The signal power, in dBm0. */
314 SPAN_DECLARE(float) v17_rx_signal_power(v17_rx_state_t *s);
315 
316 /*! Set the power level at which the carrier detection will cut in
317  \param s The modem context.
318  \param cutoff The signal cutoff power, in dBm0. */
319 SPAN_DECLARE(void) v17_rx_signal_cutoff(v17_rx_state_t *s, float cutoff);
320 
321 /*! Set a handler routine to process QAM status reports
322  \param s The modem context.
323  \param handler The handler routine.
324  \param user_data An opaque pointer passed to the handler routine. */
325 SPAN_DECLARE(void) v17_rx_set_qam_report_handler(v17_rx_state_t *s, qam_report_handler_t handler, void *user_data);
326 
327 #if defined(__cplusplus)
328 }
329 #endif
330 
331 #endif
332 /*- End of file ------------------------------------------------------------*/
void v17_rx_set_qam_report_handler(v17_rx_state_t *s, qam_report_handler_t handler, void *user_data)
Definition: v17rx.c:1563
void v17_rx_signal_cutoff(v17_rx_state_t *s, float cutoff)
Definition: v17rx.c:188
int v17_rx_restart(v17_rx_state_t *s, int bit_rate, int short_train)
Reinitialise an existing V.17 modem receive context.
Definition: v17rx.c:1380
Definition: private/v17rx.h:54
void v17_rx_set_modem_status_handler(v17_rx_state_t *s, modem_status_func_t handler, void *user_data)
Change the modem status report function associated with a V.17 modem receive context.
Definition: v17rx.c:1367
put_bit_func_t put_bit
The callback function used to put each bit received.
Definition: private/v17rx.h:59
Definition: complex.h:77
void(* put_bit_func_t)(void *user_data, int bit)
Definition: async.h:105
float v17_rx_signal_power(v17_rx_state_t *s)
Definition: v17rx.c:182
int bit_rate
The bit rate of the modem. Valid values are 7200 9600, 12000 and 14400.
Definition: private/v17rx.h:57
void(* modem_status_func_t)(void *user_data, int status)
Definition: async.h:114
float v17_rx_symbol_timing_correction(v17_rx_state_t *s)
Definition: v17rx.c:176
logging_state_t * v17_rx_get_logging_state(v17_rx_state_t *s)
Get the logging context associated with a V.17 modem receive context.
Definition: v17rx.c:1374
SPAN_DECLARE_NONSTD(int) v17_rx(v17_rx_state_t *s
Process a block of received V.17 modem audio samples.
void v17_rx_set_put_bit(v17_rx_state_t *s, put_bit_func_t put_bit, void *user_data)
Change the put_bit function associated with a V.17 modem receive context.
Definition: v17rx.c:1360
Definition: complex.h:42
int v17_rx_release(v17_rx_state_t *s)
Release a V.17 modem receive context.
Definition: v17rx.c:1550
float v17_rx_carrier_frequency(v17_rx_state_t *s)
Definition: v17rx.c:170
Definition: private/logging.h:33
int v17_rx_free(v17_rx_state_t *s)
Free a V.17 modem receive context.
Definition: v17rx.c:1556
int v17_rx_equalizer_state(v17_rx_state_t *s, complexf_t **coeffs)
Get a snapshot of the current equalizer coefficients.
Definition: v17rx.c:208
v17_rx_state_t * v17_rx_init(v17_rx_state_t *s, int bit_rate, put_bit_func_t put_bit, void *user_data)
Initialise a V.17 modem receive context.
Definition: v17rx.c:1517
int short_train
Scrambler tap.
Definition: private/v17rx.h:168